Introduction

Diabetes currently affects nearly half a billion people globally, and this number is projected to increase to 629 million by 2045, according to recent estimates by the International Diabetes Federation. The global healthcare cost of diabetes is now 727 billion US dollars per year, which is a tremendous economic burden on the society. USA spends the most on healthcare for people with diabetes, accounting for 48% of the global expenditure, followed by Europe which spends about 23% of the total expenditure (IDF diabetes atlas, eighth edition, 2017). While diabetes has already attained epidemic proportions worldwide with escalating economic costs, the prevalence of debilitating long-term complications associated with the disease such as retinopathy, nephropathy, neuropathy, and cardiovascular disease is also steadily increasing among diabetic patients. Retinopathy is the most common microvascular complication of diabetes and the chief cause of vision-impairment and blindness among working-age adults between 20 and 74 years of age (Yau et al. 2012). Risk factors associated with the development of diabetic retinopathy (DR) include hyperglycemia, dyslipidemia, and hypertension, as well as the duration of diabetes (Klein et al. 1984; Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications Research Group et al. 2000; Kohner 2008; ACCORD Study Group et al. 2010; Frank 2014; Hammer and Busik 2017). Almost all type I diabetic patients and approximately 60% of type II patients develop retinopathy after 20 years of diabetes (Aiello et al. 1998; Fong et al. 2003). Despite such a high incidence of DR among diabetic patients, the condition tends to remain undiagnosed until permanent damage to the retina has already occurred; moreover, the focus of current therapy is on treating advanced stages of DR. Understanding the pathophysiology of DR is important for developing effective and early treatment options for this devastating complication of diabetes.

Pathophysiology of Diabetic Retinopathy

DR is usually characterized by microvascular abnormalities such as irregular blood vessels and microaneurysms, which lead to microhemorrhages (Antonetti et al. 2012). Such damaged blood vessels are prone to leak blood, lipids, and fluids in the retina, resulting in macular edema and blurry vision. These exudative changes further lead to ischemic changes, such as nerve fiber damage (observed as “cotton-wool spots”), formation of microvascular collaterals or shunt vessels, and dilatation of venules. The severity of DR is diagnosed by performing fundus photography and fluorescein angiography in patients and is clinically classified as mild, moderate, or severe based on the number of hemorrhages, microaneurysms, lipoprotein exudates, and microvascular abnormalities detected. In the later proliferative stage of DR, new and delicate capillaries emerge in the retina in an attempt to compensate for the obstruction of blood flow. These abnormal new vessels are very delicate and leak readily, causing edema. They may also extend into the vitreous body, leading to vitreous hemorrhage, and this bleeding is manifested as the presence of “floaters” or dark specks in the visual field (Ivanova et al. 2016). This leads to scar tissue formation on the retina, which along with the vitreous can contract, resulting in retinal tear or detachment, causing blindness if not repaired surgically (Williams et al. 2004). Impairment of vision can also occur through retinal vein occlusion, macular edema, or neovascular glaucoma; the discussion of which is beyond the scope of this review.

Neurodegeneration in the Diabetic Retina

Although vascular abnormalities are the most visible clinical features in DR, symptoms of neuropathy such as degeneration of the inner nuclear and ganglion cell layer, and pathological changes in the nerve fiber layer have been described in the early literature on DR (Wolter 1961; Bloodworth 1962). Over the years, structural as well as functional changes in neural elements have been documented in numerous studies in animal models of DR, diabetic patients, and post-mortem human retinas (Bresnick 1986; Falsini et al. 1989; Barber et al. 1998; Lieth et al. 2000; Abu-El-Asrar et al. 2004; Antonetti et al. 2006). We now know that diabetes impacts the complete neurovascular unit of the retina, leading to neuroinflammation, gliosis or activation of glial cells, and gradual neurodegeneration that occur before the more noticeable vascular pathologies (Abcouwer and Gardner 2014). In the normal retina, the retinal vascular endothelium interacts with neuronal glia and pericytes to control retinal blood flow in response to metabolic demand: a process termed autoregulation (Kur et al. 2012; Coughlin et al. 2017). Loss of neurovascular coupling in the diabetic retina leads to impaired autoregulation and may contribute to neuronal loss due to metabolic dysfunction. The retina is composed of distinct classes of neurons: ganglion cells, amacrine cells, horizontal cells, bipolar cells, and the light-sensitive photoreceptors (Fig. 1). These neurons are organized into specific layers within the retinal tissue and form a complex neural circuitry that transmits visual signals to the brain using graded electrical activity converted to action potentials. Assessment of neuroglial function by electroretinography (ERG) indicates that ganglion, bipolar, amacrine, and photoreceptor cell functions are altered by diabetes (Barber et al. 1998; Gastinger et al. 2006; Bui et al. 2009; Kern and Berkowitz 2015; Coughlin et al. 2017). Interestingly, a recent study of type 2 diabetic patients demonstrated thinning of the inner retina by high-resolution spectral-domain optical coherence tomography (SD-OCT) (Chhablani et al. 2015). Several groups have also demonstrated that diabetes leads to rapid and early apoptosis of retinal neurons, before the onset of vascular cell death (Martin et al. 2004; Barber et al. 2005; Kern et al. 2010). Further, ERG data accurately predicts the retinal location of future microvascular damage in diabetic patients. Together, these data support the theory that the onset of neurodegeneration in the retina is an early event and not the result of vascular damage (Ng et al. 2008; Harrison et al. 2011; Simó et al. 2012). The sequence of molecular changes believed to result in retinal degeneration in diabetes is outlined in Fig. 2.

Fig. 1
figure 1

Retinal section stained with hematoxylin and eosin, showing the structure of the mouse retina and localization of various types of neurons organized within specific layers in the retina. (GCL, ganglion cell layer; IPL, inner plexiform layer; INL, inner nuclear layer; OPL, outer plexiform layer; ONL, outer nuclear layer; PL, photoreceptor layer)

Full size image
Fig. 2
figure 2

Postulated sequence of molecular changes leading to retinal degeneration in diabetes: early on in diabetes, inflammatory markers such as TNF-α, IL-1β, IL-6, MCP-1, VEGF, ICAM-1, and VCAM-1 are increased partly due to activation of glial cells, and neuroprotective factors such as NGF, BDNF, and PEDF are reduced. This leads to neuronal dysfunction, apoptosis of retinal ganglion cells, and loss of retinal synapses. Neurodegenerative changes have been observed prior to vascular pathology, which is initially manifested as loss of pericytes and endothelial cells lining retinal capillaries, gradual breakdown of the blood-retinal barrier, and increased vascular permeability. Further increase in reactivity of retinal glial cells (Müller cells, astrocytes, microglia) is observed. Upregulation of cell adhesion molecules on retinal vasculature leads to increased leukostasis and infiltration of circulating monocytes, which contribute to the vicious cycle of releasing neurotoxins and inflammatory cytokines that damages the retinal tissue. Microaneurysms, hemorrhages, leakage of lipid exudates in the retinal tissue, and loss of functional capillaries occur, leading to blurry vision. If untreated, DR can progress to the proliferative stage, involving neovascularization, macular edema, vitreous hemorrhage, and scar tissue formation, eventually causing blindness. (TNF-α, tumor necrosis factor-α; IL-1β, interleukin-1β; MCP-1, monocyte chemoattractant protein-1; VEGF, vascular endothelial growth factor; ICAM-1, intercellular adhesion molecule-1; VCAM-1, vascular cellular adhesion molecule-1; NGF, nerve growth factor; BDNF, brain-derived neurotrophic factor; PEDF, pigment epithelium-derived factor)

Full size image

Factors Associated with Retinal Neuronal Degeneration

Inflammatory Cytokines and Neurotrophic Factors

In diabetes, the integrity of the blood-retinal barrier is compromised as a result of long-term hyperglycemia, leading to loss of autoregulation, changes in permeability of endothelial cells and pericytes lining retinal capillaries, and increased secretion of cytokines and growth factors. With breakdown of the normally tight blood-retinal barrier, chronic low-grade inflammation develops in the diabetic retina. Leukocytes migrate to the retina and adhere to retinal capillaries with the help of integrins CD11b/CD18 and CD11a. The integrins bind to cell adhesion molecules (CAMs) E-selectin, ICAM-1, and VCAM-1 on the capillary endothelial cells which subsequently leads to capillary obstruction and injury; further, death of pericytes and endothelial cells results in retinal capillary loss (McLeod et al. 1995; Miyamoto et al. 1999; Barouch et al. 2000; Joussen et al. 2001). Extensive studies have demonstrated the contribution of potent inflammatory cytokines and growth factors secreted by leukocytes and activated glial cells such as TNF-α, IL-1β, IL-6, IL-8, MCP-1, and VEGF toward propagating chronic inflammation in the early stages of DR (Aiello et al. 1994; Busik et al. 2008; Abcouwer 2013).

In contrast, some pro-survival neurotrophins such as brain-derived neurotrophic factor (BDNF) are believed to protect retinal ganglion and amacrine cells from degeneration (Johnson et al. 1986). Significantly, decreased levels of BDNF have been observed in diabetic patients as well as animal models (Ola et al. 2013; Guo et al. 2017). Nerve growth factor (NGF) has been shown to prevent apoptosis of Müller glia and neurons. By contrast, its precursor protein pro-NGF is believed to promote retinal neurodegeneration by activating pro-apoptotic pathways in retinal ganglion cells (Hammes et al. 1995; Mysona et al. 2014). The oxidative milieu of the diabetic retina impairs processing of pro-NGF to mature NGF which contributes to ganglion cell death, increased vascular permeability, and inflammation (Mysona et al. 2014). Pigment epithelium-derived factor (PEDF), a potent inhibitor of angiogenesis, is believed to protect neurons from light damage and oxidative stress. Some studies have reported a decrease in PEDF levels in the diabetic retina which may contribute to inflammation and vessel leakage (Zhang et al. 2006; Yoshida et al. 2009), while supplementation with PEDF reduces retinal vascular leakage in diabetic mice (Liu et al. 2012). Ciliary neurotrophic factor (CNTF) and fibroblast growth factor (FGF) have also been reported to promote survival of retinal neurons and protect ganglion cells from degeneration in animal models, possibly by preventing diabetes-induced apoptosis of neurons (Aizu et al. 2003; Unsicker 2013; Mathews et al. 2015; Guo and Liu 2017). Figure 3 summarizes the key factors believed to be associated with neuronal degeneration in the diabetic retina.

Fig. 3
figure 3

This review provides the readers an insight into novel perspectives on diabetic retinopathy, which consider neurodegeneration an early event before the onset of more visible microvascular abnormalities in the diabetic retina. Key molecular mechanisms underlying diabetic retinal neurodegeneration are discussed, such as imbalances in neurotrophins and cytokines, glial reactivity, changes in signaling pathways and cell adhesion molecules, activation of caspases, and mitochondrial dysfunction. Understanding the molecular basis of neuronal damage is the first step toward preventing permanent retinal damage in diabetes. (BDNF, brain-derived neurotrophic factor; NGF, nerve growth factor; PEDF, pigment epithelium-derived factor; CNTF, ciliary neurotrophic factor; EPO, erythropoietin; FGF, fibroblast growth factor; TNF-α, tumor necrosis factor-α; IL-1β, interleukin-1β; MCP-1, monocyte chemoattractant protein-1; VEGF, vascular endothelial growth factor; ICAM-1, intercellular adhesion molecule-1; VCAM-1, vascular cellular adhesion molecule-1; NCAM, neural cellular adhesion molecule; AIF, apoptosis inducing factor)

Full size image

Glial Reactivity

In the healthy retina, glial cells help to maintain the integrity of the blood-retinal barrier and maintain tissue homeostasis. However, in diabetes, activation and proliferation of glia such as astrocytes, Müller cells, and microglia contributes to inflammation by secretion of pro-inflammatory mediators and neurotoxic factors (Rungger-Brändle et al. 2000). Müller cells normally help in maintaining neuronal health by recycling neurotransmitters, preventing neurotoxicity by removing excess glutamate, preventing accumulation of K+ ions by spatial buffering, participating in the retinoid cycle, and providing lactate as a fuel source for photoreceptors. Müller cell morphology is altered early on in diabetes, with enlargement of cell bodies and increased number of processes. They also demonstrate increased expression of a key marker for reactive gliosis: glial fibrillary acidic protein (GFAP) (Coughlin et al. 2017). Müller cells and astrocytes normally aid in autoregulation, provide neuronal support, and maintain ionic homeostasis in the retina. However, in early stages of diabetes, astrocyte cell numbers are decreased and their GFAP expression is markedly reduced (Rungger-Brändle et al. 2000). Although increased GFAP expression in Müller cells persists throughout DR while GFAP expression in astrocytes is seemingly downregulated, no studies have demonstrated any functional outcomes in connection with GFAP. However, these early changes in glial cells are coincident with reduced capillary density in the inner retina and functional deficits in ganglion cell responses (Ly et al. 2011).

Microglia play a critical role in initiating neural inflammatory responses. These normally dormant resident macrophages of the retina become activated in diabetes with retraction of their dendrites and contribute to neuronal damage by secreting neurotoxic factors (Colton and Wilcock 2010; Abcouwer 2013). Further, circulating monocytes are believed to contribute to inflammation by infiltrating the diabetic retina, transforming into microglia-like reactive cells, and leading to further proliferation of surrounding glial cells (Hinze and Stolzing 2011; Hu et al. 2013; Chakravarthy et al. 2016).

In the proliferative stage of retinopathy, activated glia, especially Müller cells contribute to the formation of fibrovascular scar tissue in the retina in an attempt to rebuild the injured tissue. Such gliotic scars prevent regrowth of neurons, compromise vision due to retinal detachment, and may ultimately lead to blindness (Roy et al. 2016). Major components of glial scar formation are extracellular matrix (ECM) glycoproteins such as collagen, fibronectin, chondroitin sulfate proteoglycans (CSPG), laminins, and tenascins (Ban and Twigg 2008; Barros et al. 2011). Increased synthesis and/or reduced degradation of specific ECM proteins have been shown to create an inhibitory environment that prevents axonal regeneration (Asher et al. 2001; Reinhard et al. 2017). Enzymatic digestion of CSPG reduces its inhibitory actions and promotes axonal regrowth and functional recovery in animal models of CNS injury (Bai et al. 2010; Pearson et al. 2018). These studies suggest that modulation of ECM proteins is likely to promote neuronal regeneration in the retina of diabetic individuals.

Markers of Neuronal Degeneration

Several studies have demonstrated significant reduction in the thickness of the inner plexiform layer of the retina in animal models of DR, indicating a progressive loss of neural dendrites and synapses in the inner retina (Barber et al. 1998, 2005). Structural remodeling of the ON-type ganglion cell dendrites, including changes in dendritic length and density, occurs in humans as well as animal models suggesting that a specific subset of retinal ganglia is altered by diabetes (Qin et al. 2006; Meyer-Rüsenberg et al. 2007; Gastinger et al. 2008). These pathological alterations presumably affect retinal function in diabetes, since differences in shape and size of the dendrites are believed to contribute to specificity of the visual response (Yang and Masland 1994; Brown et al. 2000). Recent studies in retinal neurons have demonstrated a critical role for CAMs in mediating cell-cell interactions, regulating dendritic morphogenesis, patterning of neuronal arbors, and maintaining synaptic specificity between the distinct types of neurons in the retina (Fuerst et al. 2008; Duan et al. 2014; Krishnaswamy et al. 2015). Focusing research attention toward CAMs may lead to better understanding of the immense diversity among retinal neuronal subtypes, which is crucial for developing neuronal regenerative therapies such as ganglion or photoreceptor cell replacement.

Neural cell adhesion molecule (NCAM), which is expressed on all neurons and glial cells of the retina, has a key role in maintaining neuronal plasticity and in preventing ganglion cell death due to retinal injury or aging (Murphy et al. 2007a). NCAM−/− mice demonstrate premature loss of visual acuity, impaired retinal function, and reduced survival of retinal neurons following injury (Luke et al. 2016a). In patients with neurodegenerative disorders such as Alzheimer’s disease, an increase in soluble NCAM is observed in serum and cerebrospinal fluid (Berezin 2010). NCAM is also believed to be involved in the modulation of pathological angiogenesis during proliferative retinopathy (Håkansson et al. 2011). These findings suggest that neuronal degeneration in the retina of diabetic individuals is likely associated with alterations in retinal and/or vitreal NCAM levels. The effect of diabetes on the function of CAMs, homophilic, and heterophilic interactions between different members of CAMs and their co-receptors on interneurons and retinal ganglia and downstream effect of these interactions on signal transduction within the neuronal cell interior are exciting new avenues that warrant further investigation.

Another set of molecular players are caspases which are known to function as central components in the apoptotic response of retinal ganglion cells. Caspases 1, 2, 6, 8, and 9 are activated as early as 8 weeks in diabetic mice (Mohr et al. 2002). Activities of caspase 1, 3, 4, 5, and 6 are increased at around 4 months after experimentally induced diabetes as well as in diabetic patients (Mohr et al. 2002; Thomas et al. 2017). Caspase 3 is regarded as a pivotal molecular player orchestrating apoptosis and bringing the cell to its demise. The pattern of caspase activation changes through the time course of diabetes and occurs as a gradual response to hyperglycemia. In general, levels of apoptotic markers such as caspase 1, 3, 6, 9, Fas, Bad, and Bax correlate highly with neuronal degeneration in DR models, as well as diabetic patients (Abu El-Asrar et al. 2007; Oshitari et al. 2010; Thomas et al. 2017).

Mitochondria play a crucial role in mediating caspase-independent apoptotic cell death, by the release of apoptogenic molecules such as apoptosis inducing factor (AIF) and cytochrome c (van Gurp et al. 2003; Schultz and Harrington 2003). Both photoreceptors and ganglion cells from diabetic patients express upregulated levels of AIF and increased immunoreactivity for cytochrome c, indicating mitochondrial dysfunction and caspase-independent apoptosis within the retina (Abu El-Asrar et al. 2007). Another molecule, a neural regeneration-associated membrane receptor expressed in retinal ganglia called Nogo receptor (NogoR), is also believed to mediate apoptosis of ganglion cells by activating the RhoA/Rock1 signaling pathway in diabetic animal models (Liu et al. 2014; Guo and Liu 2017).

Signaling Pathways in Neurodegeneration

Aberrant activation of a number of signaling cascades such as protein kinase C (PKC) and mitogen-activated protein kinase (MAPK) signaling has been implicated in structural and functional damage to retinal neurons in DR (Ahrén 2009). Increased activity of PKC-δ, an upstream regulator of the Akt cell survival pathway, was shown to contribute to retinal neuronal apoptosis in diabetic rats (Kim et al. 2008). Activation of p38 MAPK signaling has been reported in DR and is implicated in apoptosis of retinal ganglion cells, while its inhibition prevented the development of early stages of DR in a rat model (Kikuchi et al. 2000; Poulaki et al. 2002; Du et al. 2010).

Recent studies suggest a connection between G protein-coupled receptor (GPCR)-mediated signaling and diabetes (Du et al. 2015; Roy et al. 2017). Functional deterioration of G proteins, especially the inhibitory G protein (Gαi), has been observed in the diabetic rat retina (Kowluru et al. 1992). In another study, a receptor coupled to G protein, GPR91, was shown to accumulate in diabetic rat retinal ganglion cells and mediate retinal vascular damage via the MAPK signaling pathway (Li et al. 2014). The Gβγ heterodimer associates with the enzyme phosphoinositide 3-kinase (PI3K) which is an important mediator of insulin receptor signaling (Vadas et al. 2013). Furthermore, the activity of PI3K/Akt kinase pathway which is used by retinal neurons as a survival signal is reduced in diabetic rat retinas, leading to neuronal cell death (Reiter et al. 2006). GPCR-mediated regulation of neurodegenerative pathways in diabetes is a relatively new field and it is important to keep in mind the complexities of this system, such as post-translational regulation of G protein subunits which may be altered in diabetes (Kowluru 2010) or expression of additional modulatory factors that affect G protein signaling (Kowluru 2017). Some G protein isoforms may also compensate for the loss of another isoform; this was recently observed in a model of arterial thrombosis (Devanathan et al. 2015).

Current Treatment Strategies for Diabetic Retinopathy

Prevention and Control of DR

In the past, patients with diabetes had a very high risk of complete visual loss, due to a dearth of treatment options for DR, and poor control of blood glucose levels. Today, with the advent of intensive glycemic and blood pressure control, innovation in accurate monitoring of blood glucose levels, increased screening for DR and earlier diagnosis of diabetes, therapy for DR has improved dramatically. Numerous important clinical trials have demonstrated the advantages of tight control of blood glucose (with HbA1c < 7%) in the prevention of DR onset and decreasing DR progression (Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications Research Group et al. 2000; ACCORD Study Group et al. 2010). The results of the Diabetes Control and Complications Trial (DCCT) demonstrated that intensive insulin therapy significantly reduced the risk of development of DR by 75% and progression of DR by 50% after 6.5 years in type I diabetic patients, and the United Kingdom Prospective Diabetes Study (UKPDS) reported a 21% reduction in progression of DR in type II diabetic patients (Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications Research Group et al. 2000; Kohner 2008). However, highly intensive glycemic control (HbA1c of 6% or less) is not recommended since it may increase the risk of cardiovascular events and mortality due to hypoglycemia (Frank 2014).

Detection of Retinal Neurodegeneration

Clinically, DR is classified as a microvascular disease based on the observation and quantification of standard signs such as hemorrhages, microaneurysms, hard exudates, cotton-wool spots, retinal thickening, neovascularization, and other microvascular abnormalities by fundus photography, viewing only 20–70° of the retina (Wilkinson et al. 2003). However, functional, sub-clinical, and peripheral retinal changes are not incorporated in the conventional DR grading system. Novel methods such as ultra-widefield imaging allow up to 200° view of the retina in a single image, enabling correct estimation of DR severity (Kiss and Berenberg 2014). Multifocal ERG (mfERG) is another valuable technique which can be used by clinicians to predict the future location of clinical DR by mapping the precise location of visual dysfunction in the neuroretina (Ng et al. 2008; Harrison et al. 2011). This tool can help clinicians identify patients at risk of developing DR, as well as the nature and extent of retinal damage.

Typically, non-invasive techniques such as optical coherence tomography (OCT) are used for detecting total retinal thickness in cross-sectional images. Although commercial OCT devices have an axial resolution of 5–8 μm, smaller structural changes (0.5–1.0 μm) in the neuroretina can be measured using publicly available image analysis algorithms (www.iibi.uiowa.edu/content/shared-software-download) (van Dijk et al. 2011; Lynch and Abràmoff 2017). Recent developments such as the spectral-domain OCT (SD-OCT) technology allow clearer visualization of individual retinal layers with improved signal-to-noise ratio (de Boer et al. 2003). Studies using SD-OCT have demonstrated that substantial loss of neurons in patients, as indicated by reduced neuroretinal thickness, occurs very early in DR and in some cases, even in the absence of clinical DR (van Dijk et al. 2011; Chhablani et al. 2015; Sohn et al. 2016). The use of widefield technology, mfERG, and SD-OCT image analysis of the neuroretina by clinicians is essential for early detection, effective treatment, and monitoring of retinal neurodegeneration in diabetic patients.

Laser Photocoagulation

For several decades, treating patients having vision-threatening retinopathy with laser photocoagulation has been a standard therapeutic strategy for DR. The panretinal photocoagulation technique (PRP) burns the retinal tissue at several spots, sealing off leaky blood vessels and discouraging further vessel growth. This treatment has been demonstrated to reduce risk of severe impairment of sight due to abnormal vessel growth in macular degeneration and proliferative DR by approximately 50%. Although valuable, PRP may give rise to further problems like loss of peripheral vision and sub-retinal fibrosis, and it may worsen macula edema (Vander et al. 1991; Lövestam-Adrian and Agardh 2000).

Intravitreal Biologic Therapies

Recently, the management of proliferative DR has greatly benefited from a new class of biologics which directly inhibit pro-angiogenic vascular endothelial growth factor (VEGF) in the retina. Anti-VEGF antibodies, such as ranibizumab (Lucentis) and bevacizumab (Avastin), improve visual acuity in patients with macular edema and proliferative DR, as evidenced in randomized clinical trials (Nguyen et al. 2012; Brown et al. 2013; Osaadon et al. 2014; Schmidt-Erfurth et al. 2014). Despite improvements in visual acuity in the short term, the effectiveness of anti-VEGF injections for treatment of retinal diseases in the long term is still unclear. Approximately 30–40% of diabetic patients do not respond to anti-VEGF treatment, while many develop resistance over time after repeated anti-VEGF administration (Yang et al. 2016; Gonzalez et al. 2016). Follow-up studies have described complications such as increase in intraocular pressure, no significant reduction in central subfield thickness, persistence of macular edema, retinal atrophy, abnormal thinning of the retina, and suboptimal visual acuity in a substantial number of patients after 2–5 years of anti-VEGF treatment (Bressler et al. 2012; Comparison of Age-related Macular Degeneration Treatments Trials (CATT) Research Group et al. 2016; Gonzalez et al. 2016). In patients with severe proliferative DR, tractional retinal detachment may occur shortly following administration of intravitreal bevacizumab (Arevalo et al. 2008; Falavarjani and Nguyen 2013). Moreover, sustained inhibition of VEGF leads to neurodegeneration in animal models of DR, including loss of retinal ganglion, bipolar, and amacrine cells, likely due to loss of neurotrophic factors associated with VEGF depletion (Romano et al. 2012; Park et al. 2014; Hombrebueno et al. 2015). Additional data from clinical trials are needed to provide conclusive evidence regarding retinal neurodegeneration and other adverse outcomes of prolonged anti-VEGF treatment in DR.

Therapies such as intravitreal injection of corticosteroids or vitreoretinal surgery may also help reduce inflammation and prevent severe vision loss, but these approaches are associated with complications such as increased incidence of cataract and glaucoma (Diabetic Retinopathy Clinical Research Network et al. 2010; Stewart 2012).

Alternative Drug Delivery Systems

Currently, standardized long-term anti-VEGF therapy is not ideal since it requires patients to endure a high burden of healthcare visits with indefinite monthly rounds of invasive intravitreal injections (Blinder et al. 2017). It also poses a significant financial burden on patients and caregivers due to the requirement for special ophthalmic formulations, and a retina specialist to administer injections in the eye (Wallick et al. 2015). More effective methods of ophthalmic drug delivery, as well as innovative treatment strategies that target early neuroglial injury and help regenerate the retinal tissue, are clearly needed. In this context, the recent development of slow-release implants allows extended delivery of very low doses of a drug, thereby reducing many harmful side effects (Edelhauser et al. 2010). Such intravitreal implants that release dexamethasone are currently being evaluated in phase III and IV clinical trials for the treatment of diabetic macular edema, and preliminary data indicates favorable results in vision improvement (Boyer et al. 2014; Augustin et al. 2015), (ClinicalTrials.gov: NCT02471651, NCT01492400).

Topical administration of drugs such as somatostatin and glucagon-like peptide 1 (GLP-1) receptor agonists in the form of eye drops has been recently demonstrated to prevent retinal neurodegeneration in animal models of early DR (Hernández et al. 2013, 2016), and a phase II clinical trial (NCT01726075) in DR patients has shown promising initial outcomes in this direction (Simó 2017). Other routes of retinal drug delivery that are being tested in pre-clinical studies include therapeutic contact lenses containing hydrogels for molecular imprinting and mucoadhesive (poloxamer/chitosan) formulations (Gratieri et al. 2010; Hui et al. 2012).

Alternative Treatment Strategies

Apart from VEGF pathway, the angiopoietin-Tie receptor signaling pathway has been identified as an important regulator of blood-retinal barrier and vessel formation, thereby representing an alternative target for anti-angiogenic therapy in DR (Augustin et al. 2009). The concentrations of Ang-1 and Ang-2 are significantly high in animal models and DR patients, indicating an imbalance in this signaling pathway (Patel et al. 2005; Rangasamy et al. 2011; Khalaf et al. 2017). Several drugs such as MEDI3617 (AstraZeneca) and CVX-060 (Pfizer) that selectively inhibit Ang-2 (Leow et al. 2012; Wu et al. 2016), and AMG 386 (Amgen) targeting Ang1/2 are currently being evaluated in clinical trials for cancer therapy, while AMG 386 is also being tested in retinal diseases (Oliner et al. 2012; Marth et al. 2017). RG7716 (Roche) is a bispecific antibody targeting both VEGF and Ang-2, currently in phase 2 and 3 clinical trials for the treatment of diabetic macular edema (Regula et al. 2016). Sub-cutaneous injections of a Tie-2 activator, AKB-9778 (Aerpio) in combination with anti-VEGF therapy is in phase 2 clinical trials for DR and shows significantly improved visual acuity compared to anti-VEGF monotherapy (Campochiaro et al. 2016).

Retinal microvascular repair is currently being explored using stem/progenitor cell therapy. These vascular reparative cells reside in the adult bone marrow and can be harvested from peripheral blood after G-CSF administration, genetically modified ex vivo by adeno-associated virus transduction and infused intravenously to treat the same patient (Sekiguchi et al. 2009; Sen et al. 2010). These progenitor cells have the ability to home to areas of vascular injury in response to chemotactic gradients and aid in vascular repair (Vagima et al. 2011). This type of personalized treatment using stem/progenitor cells is an attractive therapeutic strategy, since these cells can potentially assist in repair of injured retinal microvasculature in DR as well as in other diabetic complications such as wound healing or limb ischemia. Additionally, intravenous delivery of progenitor cells does not involve invasive procedures or packaging into expensive ophthalmic preparations as is required for intravitreal injections, which makes this mode of therapy very attractive to patients. Currently, the safety of autologous transplantation of bone marrow stem cells for treatment is being evaluated in several phase I clinical trials (Table 1) (Park et al. 2015; Satarian et al. 2017; Cotrim et al. 2017).

Table 1 List of clinical trials currently using cell therapy/replacement therapy for treatment of retinal neurodegeneration
Full size table

Functional regeneration of retinal neurons is another promising strategy in the treatment of DR. Restoration of insulin signaling and strict glycemic control can help normalize inflammation in the diabetic retina, but they do not aid in regaining normal neuronal function (Masser et al. 2014). Although mammalian retinal cells lack the potential for spontaneous neuronal regeneration, modulation of specific proteins may help in repair of the injured retinal tissue and restoration of vision; these will be discussed in the next section.

Neuronal Regeneration in Diabetic Retinopathy

Promoters of Neuronal Growth

Cell adhesion molecules (CAMs) activate downstream signaling pathways leading to regulation of neurite outgrowth, and neuronal regeneration. Two kinds of CAMs are found on growing axons: the calcium-dependent cadherins and the immunoglobulin-superfamily CAMs (IgSF-CAMs). Interaction between these CAMs is crucial not only for forming physical attachments between cells, but also for activation of signaling cascades for growth cone navigation during neuronal development and regeneration. Cadherins have been shown to be important for regeneration of injured retinal ganglion cell axons in zebrafish and newt (Hirota et al. 2001; Liu et al. 2002). Polysialylated NCAM (PSA-NCAM) is also believed to support the survival of injured retinal ganglion cells and protect the retina from neurodegeneration (Murphy et al. 2007b, 2009; Lobanovskaya et al. 2015; Luke et al. 2016b).

Neurotrophins trigger neuroprotective signaling cascades in the neuron by interacting with receptors or integrins on the growth cone and can be used to potentially prevent neuronal degeneration. Brain-derived neurotrophic factor (BDNF) has been demonstrated to protect ganglion, glial, and amacrine cells in culture and in diabetic rats from cell death (Johnson et al. 1986; Hammes et al. 1995; Loeliger et al. 2008). Sigma 1 receptor in the retina is another promising therapeutic target for retinal degenerative diseases, and activation of this molecular chaperone has been shown to protect against ganglion cell loss and preserve cone photoreceptor function in vivo (Smith et al. 2008, 2018; Wang et al. 2016). Imbalance in levels of native neuroprotective factors in the retina such as somatostatin, insulin, PEDF, and CNTF can also be corrected by supplementation of the neurotrophin (Aizu et al. 2003; Reiter et al. 2006; Liu et al. 2012; Hernández et al. 2013). Alternatively, topical administration of a combination of neurotrophins may better protect the various types of retinal neurons and may prove more effective in preventing early retinal neurodegeneration, since intravitreal injections are not preferred in the early stages of DR when there are no visible symptoms.

Since apoptosis is the major pathway for neuronal death in DR, inhibition of apoptosis using caspase inhibitors, inhibition of microglia, and overexpressing Bcl-2 or pro-survival neurotrophic factors have been demonstrated to drastically reduce neuronal cell death in the degenerating mouse retina (Chierzi et al. 1998; Bode and Wolfrum 2003; Vigneswara and Ahmed 2016; Feng et al. 2016; Thomas et al. 2017). Interestingly, inhibition of caspases has also been demonstrated to promote limited axon regeneration in rat retinal ganglion cells (Vigneswara et al. 2014; Thomas et al. 2017).

Reprogramming Glia or RPE Cells

Reprograming other retinal cell types into functional neurons has been studied extensively in the last few years, in an attempt to recreate retinal regeneration from Müller or retinal pigmented epithelium (RPE) cells by the zebrafish and some amphibians. The RPE can potentially be reprogrammed to generate the neuroretina since RPE cells found in the chick embryo and some amphibians have a natural ability to differentiate and regenerate retinal neurons (Yoshii et al. 2007; Luz-Madrigal et al. 2014). Several studies have explored the possibility of gene-directed reprogramming of RPE cells in vitro and in vivo to differentiate toward retinal neurons (Ma et al. 2009; Li et al. 2010; Wang and Yan 2014).

Müller cells are the most abundant non-neural cell type in the retina and share a common progenitor cell lineage with photoreceptor cells. Recent studies have identified potential genes involved in reprogramming Müller cells into retinal neurons, such as Ascl1, Sox2, and Lin28 which can be modulated to stimulate functional neuronal regeneration (Pollak et al. 2013; Ueki et al. 2015; Gorsuch et al. 2017). More recently, Müller glia were successfully reprogrammed to express retinal neuronal markers and synapse with other neurons, as well as respond to light in adult mice (Jorstad et al. 2017).

Cell Replacement Strategies

The retinal ganglion cells are the most susceptible to cellular damage and neurotoxicity in response to hyperglycemia, among all neuronal types in the retina. Recent studies demonstrate that intravitreal injection of neural precursors or mature ganglion cells may help restore retinal function following neurodegeneration by integration of the transplanted neurons or progenitors in the adult retinas (Venugopalan et al. 2016; Divya et al. 2017). Once transplanted, ganglion cells must regenerate their axons in the host retina and form target-specific synapses within and beyond the eye. Supplementation with neurotrophins such as CNTF and PEDF has been demonstrated to aid axon regeneration and elongation after optic nerve injury (Zhang et al. 2016; Cen et al. 2017).

Data from ERG indicate that photoreceptors which are metabolically the most active neurons of the CNS become functionally impaired as early as 2 days after onset of diabetes in rats (Phipps et al. 2004; Kern and Berkowitz 2015). Replacement of degenerated photoreceptor cells with precursor or stem cells may improve visual function, as demonstrated by pre-clinical studies in animals (Pearson et al. 2012; Barber et al. 2013; Singh et al. 2013; Barnea-Cramer et al. 2016). Phase 1 and 2 clinical trials are ongoing for photoreceptor replacement by sub-retinal/intravitreal injection of retinal progenitor cells in patients with retinitis pigmentosa (ReNeuron, jCyte) (Table 1).

Although replacement of retinal neurons by transplantation is an attractive strategy to restore vision, the effectiveness and safety of these strategies remains to be established. Cell replacement may be hampered by the effect of the inflammatory microenvironment hosting the transplanted cells which impedes their survival and integration into the degenerating retinal tissue (Pearson et al. 2014; Singh et al. 2018). Long-term immunosuppression may be required following transplantation to prevent the immune system from destroying donor cells (Schwartz et al. 2015). Further, after in vitro expansion and transplantation of retinal progenitor cells, they must integrate into the host tissue and differentiate into the correct neuronal cell type. Although a small subset of transplanted cells derived from these cultures integrate into the host tissue and express characteristic neuronal markers, the majority of donor cells do not integrate or develop the full protein expression profile, morphology, and function of differentiated retinal neurons (Klassen et al. 2004; MacLaren and Pearson 2007; Barber et al. 2013). It is clear that major challenges lie ahead in the field of neuron replacement therapy before this technique can be applied in a clinical setting.

Other Approaches

Significant interest has been generated in developing multi-hit approaches that target neurovascular degeneration and retinal inflammation in early DR. In this context, PEDF secreted by RPE cells has multifunctional roles such as preventing neuronal dysfunction, retinal inflammation, and vascular hyperpermeability by reducing oxidative stress and glutamate excitotoxicity in the diabetic retina (Barnstable and Tombran-Tink 2004; Yoshida et al. 2009; Xie et al. 2012). Peptide derivatives of PEDF delivered in the form of eye drops have been demonstrated to reduce neurodegeneration, vascular permeability, and inflammation in retinas of diabetic mice (Liu et al. 2012). Erythropoietin (EPO) is another potent molecule that has shown promise in protecting against both neuroglial as well as vascular degeneration in DR animal models (Zhang et al. 2008; Wang et al. 2011; McVicar et al. 2011). Large, randomized clinical trials are required to confirm the potential benefits of PEDF or EPO treatment for DR in humans.

The clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9 system is an exciting new technique used for genome editing in animals as well as in cultured cells. The potential application of this revolutionary novel technology in the treatment of neurodegeneration in DR is worth exploring. Recently, this tool was used to prevent pathological angiogenesis in mouse models of retinal neovascularization by causing a 30% reduction in the expression of VEGFR2 gene in vascular endothelial cells in the mouse retina (Huang et al. 2017).

Other approaches include the use of surgically implanted electronic retinal prosthetics, transplantation of whole retinal sheets, viral expression of light-sensitive rhodopsin, and the use of synthetic photoswitches to restore a portion of the neuronal activity within the injured retina (Busskamp et al. 2010; Yue et al. 2016; Seiler et al. 2017; Tochitsky et al. 2017).

Future Directions

In order to accurately evaluate the feasibility of the above-mentioned treatment strategies for early retinal neurodegeneration in diabetes, the availability of robust animal models of DR is vital. Mice and rats are the most commonly used animal models, but dogs, pigs, rabbits, and zebrafish are also used to study the pathogenesis of DR (Robinson et al. 2012; Olivares et al. 2017). Most DR animal models whether generated via drug or diet-induced diabetes, galactosemia, or genetic models tend to display early vascular pathological changes such as pericyte loss, formation of microaneurysms, and basement membrane thickening (Kern and Engerman 1996; Agardh et al. 1997). However, none of these DR models completely mimics all the vascular and neuronal changes that occur in each stage of DR in humans (Lai and Lo 2013; Olivares et al. 2017). Type 1 DR rodent models such as streptozotocin-induced diabetes and some genetic models (NOD, Ins2Akita, db/db mice), show accelerated ganglion cell loss, distinct ONL, INL, and IPL thinning, and changes in glial responses indicating progressive neuronal damage in the diabetic retina (Rungger-Brändle et al. 2000; Martin et al. 2004; Cheung et al. 2005; Gastinger et al. 2008; Liu et al. 2012). Most of these studies have used type 1 diabetic animal models due to the lack of well-characterized type 2 diabetic animal models. A recent characterization of the ob/ob mouse model of type 2 diabetes shows promise as a model of early neurodegenerative changes in the diabetic retina (Lee et al. 2018). Since the majority of DR patients are type 2 diabetic, more future studies on early retinal neurodegeneration in type 2 DR models are warranted.

In summary, DR is a complication of diabetes that appears to involve neuropathy as well as microvasculopathy. Although historically, vision loss in diabetic patients has been attributed to microvascular damage to the retina, recent studies reveal that DR is a neurodegenerative disorder, with evidence strongly pointing to the occurrence of neuronal damage prior to overt vascular dysfunction (Sohn et al. 2016; Lynch and Abràmoff 2017). Several clinical trials have been initiated to test the safety and feasibility of using cell replacement therapies, stem cell transplantation, or the use of sub-retinal implants for the treatment of retinal neurodegeneration (Table 1).

Since the combination of retinal inflammation, neurodegeneration, and microvascular injury together contributes to the clinical signs of DR, treatment of DR should ideally aim to be three-fold: (1) addressing retinal inflammation, (2) functional regeneration of retinal neurons, and (3) repair of damaged retinal vasculature. Further studies are warranted to evaluate whether such combination therapies represent a viable choice as opposed to monotherapy and ultimately improve visual function in DR patients. Additionally, reliable biomarkers are needed to detect the extent of early retinal neurodegeneration and identify which patients would benefit the most from neuroprotective/regenerative treatment.